Cosmic Ray rejection with attention augmented deep learning

Cosmic Ray (CR) hits are the major contaminants in astronomical imaging and spectroscopic observations involving solid-state detectors. Correctly identifying masking them is a crucial part of image processing pipeline, since it may otherwise lead to spurious detections. For this purpose, we have developed tested novel Deep Learning based framework for automatic detection CR from data two different imagers: Dark Energy Camera (DECam) Las Cumbres Observatory Global Telescope (LCOGT). We considered baseline models namely deepCR Cosmic-CoNN, which current state-of-the-art learning algorithms that were trained using Hubble Space (HST) ACS/WFC LCOGT Network images respectively. experimented with idea augmenting Attention Gates (AGs) improve performance. our on DECam demonstrate consistent marginal improvement by adding AGs True Positive Rate (TPR) at 0.01% False (FPR) Precision 95% TPR over aforementioned dataset. proposed AG augmented provide significant gain FPR when previously unseen LCO test having three distinct telescope classes. Furthermore, without attention augmentation outperform such as Astro-SCRAPPY, Maximask (that natively data) pre-trained ground-based Cosmic-CoNN. This study demonstrates module enables us get better Cosmic-CoNN their generalization capability data.